AI Digest by Angtiger

2026-02-17 22:41 KST Β· by Angtiger Β· 맀일 10:00 KST μ—…λ°μ΄νŠΈ
5 sources
51 new posts
πŸ” 검색 Β· πŸ“’ TelegramAI DigestDEX Digest Β· 🌐 DigestAI DigestDEX Digest Β· πŸ“§ λ¬΄λ£Œκ΅¬λ…
πŸ“§ λ¬΄λ£Œκ΅¬λ…ν•˜κΈ°

πŸ† AI λͺ¨λΈ 벀치마크

β–Ό

πŸ–₯️ Terminal-Bench 2.0 (Top 5)

Source: OpenAI, Anthropic, Terminal-Bench

πŸ† Chatbot Arena ELO (Top 5)

Source: Chatbot Arena

🧠 ARC-AGI-2 달성λ₯ 

84.6%
πŸ€– 84.6% β€” Gemini 3 Deep Think (Google) πŸ§‘ Human Panel = 100% κΈ°μ€€
Source: ARC Prize Leaderboard
← 전체 보기 πŸ“‚ Company News×
총 51건 Β· νŽ˜μ΄μ§€ 2/3
26.02.03 Hugging Face Blog Training Design for Text-to-Image Models: Lessons from Ablations
β–² 0
Photoroom의 Text-to-Image λͺ¨λΈ ν•™μŠ΅ 섀계: Ablation μ‹€ν—˜μ—μ„œ 얻은 κ΅ν›ˆ.
26.02.03 Hugging Face Blog The Future of the Global Open-Source AI Ecosystem: From DeepSeek to AI+
β–² 0
DeepSeekλΆ€ν„° AI+κΉŒμ§€, κΈ€λ‘œλ²Œ μ˜€ν”ˆμ†ŒμŠ€ AI μƒνƒœκ³„μ˜ λ―Έλž˜μ— λŒ€ν•œ Hugging Face λΈ”λ‘œκ·Έ.
26.02.03 Hugging Face Blog H Company's new Holo2 model takes the lead in UI Localization
β–² 0
H Company의 Holo2 λͺ¨λΈμ΄ UI Localization λΆ„μ•Όμ—μ„œ 선두λ₯Ό μ°¨μ§€ν–ˆλ‹€.
26.01.29 Hugging Face Blog Introducing Daggr: Chain apps programmatically, inspect visually
β–² 0
Daggr μ†Œκ°œ: 앱을 ν”„λ‘œκ·Έλž˜λ° λ°©μ‹μœΌλ‘œ μ²΄μ΄λ‹ν•˜κ³  μ‹œκ°μ μœΌλ‘œ κ²€μ‚¬ν•˜λŠ” 도ꡬ.
26.01.29 Google DeepMind Project Genie: Experimenting with infinite, interactive worlds
β–² 0
Google AI Ultra κ΅¬λ…μž(λ―Έκ΅­)κ°€ λ¬΄ν•œν•˜κ³  μΈν„°λž™ν‹°λΈŒν•œ 세계λ₯Ό μƒμ„±ν•˜κ³  νƒν—˜ν•  수 μžˆλŠ” μ‹€ν—˜μ  연ꡬ ν”„λ‘œν† νƒ€μž… Project Genieλ₯Ό μ²΄ν—˜ν•  수 μžˆλ‹€.
26.01.28 Hugging Face Blog We Got Claude to Build CUDA Kernels and teach open models!
β–² 0
Claudeλ₯Ό ν™œμš©ν•˜μ—¬ CUDA 컀널을 λΉŒλ“œν•˜κ³  μ˜€ν”ˆ λͺ¨λΈμ— κ°€λ₯΄μΉ˜λŠ” 방법.
26.01.27 Hugging Face Blog Unlocking Agentic RL Training for GPT-OSS: A Practical Retrospective
β–² 0
LinkedIn의 GPT-OSSλ₯Ό μœ„ν•œ μ—μ΄μ „νŠΈ RL ν•™μŠ΅ μ‹€μ „ 회고둝.
26.01.27 Hugging Face Blog Alyah ⭐️: Toward Robust Evaluation of Emirati Dialect Capabilities in Arabic LLMs
β–² 0
TII의 Alyah: μ•„λžμ–΄ LLMμ—μ„œ 에미라티 λ°©μ–Έ λŠ₯λ ₯을 κ²¬κ³ ν•˜κ²Œ ν‰κ°€ν•˜κΈ° μœ„ν•œ 연ꡬ.
26.01.27 Hugging Face Blog Architectural Choices in China's Open-Source AI Ecosystem: Building Beyond DeepSeek
β–² 0
DeepSeek을 λ„˜μ–΄μ„  쀑ꡭ μ˜€ν”ˆμ†ŒμŠ€ AI μƒνƒœκ³„μ˜ μ•„ν‚€ν…μ²˜ 선택에 λŒ€ν•œ Hugging Face 뢄석.
26.01.21 Hugging Face Blog AssetOpsBench: Bridging the Gap Between AI Agent Benchmarks and Industrial Reality
β–² 0
IBM Research의 AssetOpsBench: AI μ—μ΄μ „νŠΈ λ²€μΉ˜λ§ˆν¬μ™€ μ‚°μ—… ν˜„μ‹€ κ°„μ˜ 격차λ₯Ό ν•΄μ†Œ.
26.01.16 Google DeepMind D4RT: Teaching AI to see the world in four dimensions
β–² 0
4D μž₯λ©΄ 볡원과 좔적을 μœ„ν•œ 톡합 AI λͺ¨λΈ D4RT μ†Œκ°œ. AIμ—κ²Œ 세계λ₯Ό 4μ°¨μ›μœΌλ‘œ λ³΄λŠ” 법을 κ°€λ₯΄μΉœλ‹€.
26.01.13 Google DeepMind Veo 3.1 Ingredients to Video: More consistency, creativity and control
β–² 0
Veo 3.1 μ—…λ°μ΄νŠΈ: μžμ—°μŠ€λŸ½κ³  역동적인 클립 생성 및 μ„Έλ‘œ μ˜μƒ 지원. 더 λ§Žμ€ 일관성, μ°½μ˜μ„±, μ œμ–΄λ ₯을 μ œκ³΅ν•œλ‹€.
25.12.23 Google DeepMind Google's year in review: 8 areas with research breakthroughs in 2025
β–² 0
Google 2025λ…„ 리뷰: 8개 λΆ„μ•Όμ˜ 연ꡬ 돌파ꡬ μš”μ•½.
25.12.17 Google DeepMind Gemini 3 Flash: frontier intelligence built for speed
β–² 0
Gemini 3 Flash: λΉ„μš©μ˜ μΌλΆ€λ‘œ μ΅œμ²¨λ‹¨ 지λŠ₯κ³Ό 속도λ₯Ό μ œκ³΅ν•˜λŠ” ν”„λ‘ ν‹°μ–΄ λͺ¨λΈ.
25.12.16 Google DeepMind Gemma Scope 2: helping the AI safety community deepen understanding of complex language model behavior
β–² 0
Gemma 3 μ œν’ˆκ΅°μ„ μœ„ν•œ 포괄적 μ˜€ν”ˆ 해석가λŠ₯μ„± 도ꡬ λͺ¨μŒ Gemma Scope 2 λ°œν‘œ. AI μ•ˆμ „ 연ꡬλ₯Ό κ°€μ†ν™”ν•˜κΈ° μœ„ν•΄ λ³΅μž‘ν•œ μ–Έμ–΄ λͺ¨λΈ λ™μž‘μ˜ 이해λ₯Ό λ•λŠ”λ‹€.
25.12.12 Google DeepMind Improved Gemini audio models for powerful voice experiences
β–² 0
κ°œμ„ λœ Gemini μ˜€λ””μ˜€ λͺ¨λΈλ‘œ κ°•λ ₯ν•œ μŒμ„± κ²½ν—˜μ„ μ œκ³΅ν•œλ‹€.
25.12.11 Google DeepMind Deepening our partnership with the UK AI Security Institute
β–² 0
Google DeepMind와 영ꡭ AI Security Institute(AISI)κ°€ μƒˆλ‘œμš΄ 연ꡬ νŒŒνŠΈλ„ˆμ‹­μ„ 톡해 AI μΆ”λ‘  λͺ¨λ‹ˆν„°λ§, μ‚¬νšŒκ²½μ œμ  영ν–₯ 평가 λ“± 핡심 μ•ˆμ „ 연ꡬ λΆ„μ•Όμ—μ„œ ν˜‘λ ₯을 κ°•ν™”ν•œλ‹€.
25.12.09 Google DeepMind FACTS Benchmark Suite: Systematically evaluating the factuality of large language models
β–² 0
FACTS Benchmark Suite: LLM의 사싀성을 λ§€κ°œλ³€μˆ˜, 검색, λ©€ν‹°λͺ¨λ‹¬ μΆ”λ‘  3개 μ˜μ—­μ—μ„œ μ²΄κ³„μ μœΌλ‘œ ν‰κ°€ν•˜λŠ” 벀치마크.
25.12.04 Google DeepMind Engineering more resilient crops for a warming climate
β–² 0
κ³Όν•™μžλ“€μ΄ AlphaFoldλ₯Ό μ‚¬μš©ν•˜μ—¬ κ΄‘ν•©μ„± νš¨μ†Œ(GLYK)λ₯Ό κ°•ν™”ν•˜κ³ , μ˜¨λ‚œν™”μ— 적응할 수 μžˆλŠ” 더 회볡λ ₯ 있고 λ‚΄μ—΄μ„± μžˆλŠ” μž‘λ¬Όμ„ κ°œλ°œν•˜λŠ” 길을 μ—΄κ³  μžˆλ‹€.
25.11.25 Google DeepMind Revealing a key protein behind heart disease
β–² 0
κ³Όν•™μžλ“€μ΄ AlphaFoldλ₯Ό μ‚¬μš©ν•˜μ—¬ 심μž₯λ³‘μ˜ 핡심 λ‹¨λ°±μ§ˆμ„ λ§€ν•‘ν•˜κ³ , 이 λŒνŒŒκ΅¬κ°€ μΉ˜λ£Œλ²•μ„ μ–΄λ–»κ²Œ λ³€ν™”μ‹œν‚¬ 수 μžˆλŠ”μ§€ νƒκ΅¬ν•œλ‹€.
Β«β€Ή 이전 1 2 3 λ‹€μŒ β€ΊΒ»